TiCl4
Chemical
- May 1, 2019
- 631
The previous thread ( thread124-464517 ) is closed, so I'd like to give an update on this for those who are interested and helped out. This may be helpful for anyone who has to design thickeners for light or heavy particles in the future.
We trialed a low-speed disc-stack centrifuge. Unfortunately, the particles separated too easily at low speed and agglomerated, plugging off the discs.
I was unable to find any other centrifuge types available for trials, so I decided to do some experimentation on my own to see if the thickener design could be optimized. I hope the process I describe below may be helpful for someone who needs to design thickeners for new mixtures.
The solids separation is something that can be visually seen (i.e. there is a visible "phase" boundary between the heavy water phase and the floating polymer/water phase), so I set up 1L bottles in an oven, shook them all, and measured height of the boundary layer versus time. This gave me a data set of position versus time. I then used Minitab to regress this dataset (non-linear). The best fit I found was of the form:
Position = A*t/(B+t), t=time
Taking the derivative of this equation, I got a function of boundary layer rising velocity versus time in the form of
V = C/(D+t)^2
By plugging time of original height measurements back into that equation, I could find the approximate settling velocity at the moments I made my original height measurements. This gave a data set of velocity versus settling height. I also measured the solids remaining in the bottom water water, so a mass balance let me convert settling height to a %solids value. I know there is a solids% gradient in the floating layer here, but the process will re-mix the concentrated solids later, so only average solids% matters here.
From this, I had a data set of settling velocity versus average solids%. Running a curve fit to this data, I finally had an equation y = f(x), where y = settling velocity and x=average solids. This equation I directly used to find settling times for various vessel sizes by calculating settling velocity at time 0, 15, 30, etc. and recalculating solids at each step to adjust settling velocity. The increments are small enough to give a near-continuous curve.
Finally, I validated this model, first against a 250-gal settling tank and finally against a 4,500 gallon tank, and it predicted well in both cases. From this, I designed a final thickener tank that we will be installing.
I know many applications for settling are continuous and deal with sinking solids, but I can see this approach working, with slight modifications, for developing necessary models for those situations as well (as you will have to predict settling velocity in the free and hindered settling zones on a continuous basis).
I hope this description of the general process I used was helpful, or at least interesting, to those who may be in a similar situation in the future.
We trialed a low-speed disc-stack centrifuge. Unfortunately, the particles separated too easily at low speed and agglomerated, plugging off the discs.
I was unable to find any other centrifuge types available for trials, so I decided to do some experimentation on my own to see if the thickener design could be optimized. I hope the process I describe below may be helpful for someone who needs to design thickeners for new mixtures.
The solids separation is something that can be visually seen (i.e. there is a visible "phase" boundary between the heavy water phase and the floating polymer/water phase), so I set up 1L bottles in an oven, shook them all, and measured height of the boundary layer versus time. This gave me a data set of position versus time. I then used Minitab to regress this dataset (non-linear). The best fit I found was of the form:
Position = A*t/(B+t), t=time
Taking the derivative of this equation, I got a function of boundary layer rising velocity versus time in the form of
V = C/(D+t)^2
By plugging time of original height measurements back into that equation, I could find the approximate settling velocity at the moments I made my original height measurements. This gave a data set of velocity versus settling height. I also measured the solids remaining in the bottom water water, so a mass balance let me convert settling height to a %solids value. I know there is a solids% gradient in the floating layer here, but the process will re-mix the concentrated solids later, so only average solids% matters here.
From this, I had a data set of settling velocity versus average solids%. Running a curve fit to this data, I finally had an equation y = f(x), where y = settling velocity and x=average solids. This equation I directly used to find settling times for various vessel sizes by calculating settling velocity at time 0, 15, 30, etc. and recalculating solids at each step to adjust settling velocity. The increments are small enough to give a near-continuous curve.
Finally, I validated this model, first against a 250-gal settling tank and finally against a 4,500 gallon tank, and it predicted well in both cases. From this, I designed a final thickener tank that we will be installing.
I know many applications for settling are continuous and deal with sinking solids, but I can see this approach working, with slight modifications, for developing necessary models for those situations as well (as you will have to predict settling velocity in the free and hindered settling zones on a continuous basis).
I hope this description of the general process I used was helpful, or at least interesting, to those who may be in a similar situation in the future.